OPTIMAL DESIGN OF BALL BEARING RETAINERS USING TAGUCHI METHODS AND BEARING DYNAMIC ANALYSIS

Kazuaki Maniwa(1), Takashi Nogi(1), Kazuo Natori(1) & Shingo Obara(1) Japan Aerospace Exploration Agency (JAXA) Tsukuba Space Center, 2-1-1 Sengen, Tsukuba-shi, Ibaraki-ken 305-8505, JAPAN Email: maniwa.kazuaki@jaxa.jp ABSTRACT The authors developed an optimal design technique for ball bearing retainers, which combines the Taguchi method and a bearing dynamic analysis and takes a mixed lubrication model for retainer-race and retainerball contacts into consideration. It is clarified th at retainer stability under bearing rotation can be improved by designing a high signal-tonoise ratio using the present design technique. The designed retainer with a high signal-to-noise ratio performed well in a b earing rotating test, in which no significantly increased torque and audible noise were observed. 1. INTRODUCTION their mating surfaces. Therefore, designing the retainer clearances appropriately is vital to prevent retainer instability. However, hardly any design approaches taking the optimization technique into account have been conducted to d ate, with designers mostly reliant on parametric experiments and lessons learned. To clarify the g eneration mechanism of retainer instability and design the ball bearing retainer numerically, the authors developed a bearing dynamic analysis model, which takes a mixed lubrication model for the retainer-race and retainer-ball contacts int o consideration [1], [2]. The retainer design is lik ely to become more efficient by combining the bearing dynamic analysis with an optimization technique. The objective of this study is to develop an optimal design technique for ball bearing retainers using the parameter design technique in the Taguchi method and bearing dynamic analysis. The Taguchi method is generally thought to be capable of esti mating “robustness”, which is probably suitable for evaluating retainer stability. 2. BEARING DYNAMIC ANALYSIS
(1)

Oil-lubricated ball bearings used for space applications such as a gy roscope and reaction wheel are us ually operated at rel atively high rotational speeds, typically thousands of RPM, and must also rotate with extremely low frictional torque under a vacuum. Therefore, the amount of lubricant oil must be strictly restricted to reduce the v iscous friction of oil. A cotton-based phenol retainer impregnated with the lubricant oil has been used to achieve a long operating lifetime in minimally lubricated ball bearings. During the operation of ball bearings assembled with a cotton-based phenol retainer, retainer instability often leads to probl ematic increased frictional torque and audible noise. According to the author’s experiences, the incidence of retainer instability is g reatly affected by the dimensions of guiding clearance and p ocket clearances in Fig. 1 and frictional characteristics at
Outer race Guiding clearance Pocket axial clearance ball

In this analysis, the outer race is assumed to be fixed in space, while the inner race rotates at c onstant speed. Equations of the motion of balls and retainer are numerically integrated, considering the gravity in 6degrees of fre edom. During ball-race contact, normal load is calculated using the Hertz equation, whereupon traction force and rolling resistance are estimated. For retainer-race and retainer-ball contacts, normal load and friction force are calculated according to the mixed lubrication model, whereby the oil film thickness and surface roughness of the retainer are considered [1], [2]. The normal load and friction force between the retainer and race/ball are expressed as:

w  w f  wa f  f f  fa

(1) (2)

Inner race

Retainer

Pocket circumferential clearance

Figure 1. Retainer clearances in case of inner race guiding and rectangular pocket
_________________________________________________ ‘14th European Space Mechanisms & Tribology Symposium – ESMATS 2011’ Constance, Germany, 28–30 September 2011

where w and f denote contact load and friction force, respectively and subscripts f and a denote fluid film lubrication and asperity contact, respectively.

67

096 [Pa·s] : corresponding to the viscosity of Multiply Alkylated Cyclopentane at 40 ºC Bearing attitude Horizontal shaft in 1 G Simulation time 6 [s] (corresponding to 500 revolutions of the inner race) control factor to evaluate its effect on retainer stability. A horiz ontal shaft at 1 G for the bearing attitude is app lied to simulate tests. the factors and levels of which are shown in Tab. Numerical analysis conditions Bearing type Angular contact ball bearing Bearing size Inner diameter 10 [mm] Outer diameter 26 [mm] Width 8 [mm] Number of balls 9 Rotational speed 5000 [RPM] of inner race Lubricant viscosity 0. Table 1. we identify the system output associated with the design concept and set control factors and noise factors for a robust design. As shown in Tab.1  thickness F Pocket oil film 0. 2.9 thickness Table 4. 4 was selected. the mean of th e retainer translational velocity is assumed to be an output in the Taguchi method to eva luate the retainer stability. 3.1 0. the oil film thickness is se lected as a H 1 2 3 3 1 2 3 1 2 1 2 3 2 3 1 2 3 1 68 . Noise factors result in the output variation of the system used to evaluate robustness. 1.1 (3) where subscript x corresponds to axial direction and subscripts y and z correspond to the radial direction in an inertial coordinate system. In this study. APPLICATION OF THE TAGUCHI METHOD In the parameter design of the Taguchi method. M: middle. a total of 72 ( = 18 × 4 noise factors) cases of numerical analyses were conducted. The retainer translational velocity v is expressed as 2 2 v  vx  vy  v z2 Table 3. In the present design technique. For exam ple. the oil film thickness and t he surface roughness of the retainer are chosen as control factors. outer diameter of 26 mm and width of 8 mm respectively. Control factors Control factors A Preload B Guiding clearance C Pocket axial clearance D Pocket circumferential clearance E Guiding land oil film thickness F Pocket oil film thickness G Guiding land surface roughness H Pocket surface roughness S: small. slight changes in the oil film thicknesses were gi ven to the gui ding land and t he pockets as noise factors. Orthogonal array L18 No.9 1. L: large 1 S S S S S S S S Level 2 L M M M M M M M 3 L L L L L L L 3.The numerical analysis conditions are listed in Tab. because the retainer translational velocity is sen sitive to whether retainer instability occurs. The oil film thickness is usually determined by the initial amount of supplied oil and operating conditions such as rotational speed and preload. T o investigate eight kinds of control factors. The retainer-race and the retainer-ball clearances. The bearing is assumed to be lub ricated with Multiply Alkylated Cyclopentane at a tem perature of 40 ºC. We selected eight parameters as Taguchi method control factors. with an inner diameter of 10 mm. an orthogonal array L18 shown in Tab. because a rectangular pocket shape was presumed to be more advantageous in terms of stabilizing retainer behavior than the circular pocket shape already reported in [1].9 times the nominal level corresponds to N1. whereupon the si gnal-to-noise (S/N) ratio and sensitivity in the case o f nominal-the-best Table 2. the state where both oil film thicknesses decrease to 0. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 A 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 B C 11111 12222 13333 21122 22233 23311 31213 32321 33132 11332 12113 13221 21231 22312 23123 31323 32131 33212 D E F G 1 2 3 3 1 2 2 3 1 2 3 1 3 1 2 1 2 3 N4 1. The targeted bearing is an angular contact ball bearing. A rectangular pocket shape was selected in this study. Noise factors Noise factors N1 N2 N3 E Guiding land oil film 0.1 1. Namely.9 1. on th e premise that the motion of the retainer is prone to fluctuate compared with ze ro gravity conditions [3].

04 0.81 -15.16 0. whereby the optimal condition shows the extent to which the S/N ratio and sensitivity have improved relative to the related condition.06 0. the smaller the retainer translational velocity. Conversely. In addition. The high retainer translational velocity at number 10 is due to the retainer instability. th e pocket oil film thickness (F) an d the pocket surface roughness (H).97 Confirmation S/N Sensitivity ratio 36. The prediction and confirmation of the gain Prediction S/N Sensitivity ratio 49. respectively. Table 4.18 Mean retainer translational velocity [m/s] 0.08 0. (2) The lower th e sensitivity. namely the greater the “robustness”. 6 (a).12 0. Gain here refers to the reference value. 5. The related condition is assu med to be the worst condition of t he S/N ratio in the presen t design technique. if the results of confirmation correspond to the prediction in the gai n. The interpretations of th e S/N ratio and sensitivity are as follows: (1) The higher the S/N ratio. th e sensitivity can remain small when the S/N ratio is designed to be high.85 -39. namely “good repeatability”. 4 shows the levels of each control factor with the S/N ratio under optimal and worst co nditions. Numerical analysis results of the mean retainer translational velocity 32 Signal-to-noise ratio [dB] 30 28 26 24 22 20 18 A A 1 2 B B B 1 2 3 C C C 1 2 3 D D D 1 2 3 E E E 1 2 3 F F F 1 2 3 G G G 1 2 3 H H H 1 2 3 Figure 3. The reason why a difference between the prediction and confirmation in the S/N ratio is generated can be explained by (1) interaction between the retainer clearances and the oil film thicknesses.90 46.54 -0. The numbers in th e horizontal axis correspond to each analysis condition in the orthogonal array L18.13 -11.47 -32.1 0. Optimal and worst conditions of the S/N ratio A B C D E F G H Optimal 1 31 1 3 113 condition Worst 2 13 3 2 321 condition Table 5. the prediction and confirmation differ by 9.90 Figure 4.95 -23. whereas the retainer translational velocity at N2 of number 15 differs significantly from other noise 0. 4. Basically. the retainer translational velocity is l ow and stab le as shown in Fig. and (2) the occurrence of retainer instability. 5 shows results o f the prediction and confirmation of the gains. NUMERICAL ANALYSIS RESULTS conditions. 2 shows the numerical analysis results of the mean retainer translational velocity. Graphs of the factorial effects of the sensitivity Figs. 4. Tab. namely poor “robustness”. Graphs of the factorial effects of the S/N ratio -30 -31 Sensitivity [dB] -32 -33 -34 -35 -36 -37 -38 AA 1 2 B B B 1 2 3 C C C 1 2 3 D DD 1 2 3 E E E 1 2 3 F F F 1 2 3 G G G 1 2 3 H H H 1 2 3 Optimal condition Worst condition Gain 37. In general. 5 and 6 show the numerical analysis results of the retainer translational velocity and loci of the retainer center under optimal and worst conditions respectively. the parameter that changes the S/N ratio in particular is the pocket oil film thickness (F) and the pocket surface roughness (H).66 -20. 5 (a). under the worst condition. Under the optimal condition. the retainer translational velocity is high and unstable as shown in 69 . as sh own in Fig.characteristic were calculated from the num erical analysis results of 72 cases. 3 and 4 show graphs of the factorial effects of the S/N ratio and the sensitivity.02 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 N1 N2 N3 N4 Figure 2. the less the noise factors influence the retainer translational velocity.14 0. and the retainer shows pendulum-like motion at the bottom of the clearance circle according to the gravity in Fig. 3. The S/N ratio rises when the guiding clearance (B) increases and both pocket clearances (C and D) decline.64 Fig. the analysis re sults can be implemented with high reliability. the parameter that particularly changes the sensitivity is the preload (A).78 2. As for the S/N ratio gain. From Fig.8 dB from Tab. Tab. Figs. Conversely.

7 shows the numerical analysis results of the frictional torque under optimal and worst conditions respectively. 2. Numerical analysis results of the retainer translational velocity Guiding clearance circle Gravity (a) Optimal condition (b) Worst condition Figure 6. It is pre sumed from Fig. 5 (b). 2.5 0. 6 (b). Test Method The test bearing is an angular contact ball bearing with inner diameter of 10 mm. BEARING ROTATING TEST 0. 5. 3 that the S/N ratio of retainer type A is h igher than that of retainer type B. and the retainer shows whirling motion in Fig. Numerical analysis results of the frictional torque 5. 5-7 are ju dged to indicate that the present design technique is effective in stabilizing the retainer behavior and improving the bearing frictional torque. Two kinds of cotton-based phenol retainers were prepared: (1) type A.1 0 0 50 100 150 200 Number of rotations of inner race [rev.3 0.Fig. 7 (b) is thought to be attributable to the increased contact load applied to the retainer-race and retainer-ball. induced by retainer instability.4 0. Fig. A pai r of test bearings set up i n a vacuum chamber i s rotated by the motor through the magnetic coupling. which is thou ght to indicate retainer instability.6 0.8 dB in th e S/N ratio). An axial load of ab out 29 N is applied to the test bearings by position preloading.] 250 3 2 1 0 -1 -2 -3 -4 0 50 100 150 200 250 Number of rotations of inner race [rev.] (a) Optimal condition 4 Frictional torque [mN-m] 0. The (b) Worst condition Figure 5.1 0 0 50 100 150 200 Number of rotations of inner race [rev. th e pocket circumferential clearance of which is sligh tly larger than level 3 in Tab. The guiding and pocket axial clearances in types A and B are set to be level 2. Fig. th e numerical analysis results in Fig s.5 0.2 0.1. Retainer translational velocity [m/s] 4 Frictional torque [mN-m] 3 2 1 0 -1 -2 -3 -4 0 50 100 150 200 250 Number of rotations of inner race [rev. Although insufficient repeatability is obtained (variation of 9. Numerical analysis results of loci of the retainer center 70 .6 0.2 0.] 250 To examine the stability of the designed retainer. (2) typ e B. The increased fluctuation in frictional torque in Fig.] (a) Optimal condition Retainer translational velocity [m/s] (b) Worst condition Figure 7. the pocket circumferential clearance of which is slightly s maller than level 1 in Tab. a bearing rotating test was conducted. Moreover. of equivalent size to that used for the numerical analysis.4 0. 8 sh ows a photograph and schematic diagram of the bearing tester.3 0. outer diameter of 26 mm and width of 8 mm. their surface roughness and oil film thickness were considered almost similar. since retainer types A and B were fabricated and lubricated under equivalent conditions.

Bearing tester Table 6. respectively. low temperature (-4 ºC). 80 . 24 hr (c) Vacuum pressure (less than 100 Pa). low temperature Load cell Temperature 60 40 20 0 (b) Schematic diagram Figure 8. 2000. whereupon the bearing rotating test at 5000 RPM was started.Temperature 60 40 20 0 Frictional torque [mN-m] -10 8 6 4 2 0 0 4 8 12 Frictional torque 16 20 24 Time [hr] (a) Atmospheric pressure. The bearing rotating tests were conducted at a rotational speed of 5000 RPM in three different environments. the bearings were operated at rotational speeds of 1000. temperature (22 ºC). 9 Results of bearing rotating test for retainer type A with high S/N ratio 71 Bearing temperature [ºC] 80 Bearing temperature [ºC] 80 Bearing temperature [ºC] frictional torque and temperature of the test bearings under rotation are measured with the load c ell and thermocouple. nitrogen gas). over 90 hr Frictional torque [mN-m] 8 6 4 2 0 0 20 40 60 80 100 120 -10 Frictional torque Time [hr] (c) Vacuum pressure. room temperature A pair of test bearings 60 Temperature 40 20 0 (a) Photograph Vacuum chamber Cooling shroud Frictional torque [mN-m] -10 8 6 4 2 0 0 4 8 12 16 20 24 A pair of test bearings Magnetic coupling Motor Frictional torque Time [hr] (b) Atmospheric pressure. the test bearings can be cooled to about -4 ºC. 3000 and 4000 RPM ev ery 10 min. Tab. because the high lubricant viscosity meant retainer instability was con sidered likely to occur. The low temperature environment (-4 ºC) was selected as the test environment. room temperature. By circulating an ant ifreeze solution through the cooling shroud. room temperature (22 ºC). 24 hr test duration) (b) Atmospheric pressure (dry nitrogen gas). 6 shows the bearing rotating test co nditions. room temperature Figure. Bearing rotating test conditions Rotational speed 5000 [RPM] of the inner race Lubricating oil Multiply Alkylated Cyclopentane (formulated) Environment (a) Atmospheric pressure (dry (pressure. During the initial stage of th e test.

the 1. the retainer translational velocity is selected as the output. & Obara. CONCLUSIONS Temperature 20 0 Frictional torque [mN-m] -10 8 6 4 2 0 0 4 8 12 16 20 24 Frictional torque The optimal design technique for a ball bearing retainer was developed combining bearing dynamic analysis and the parameter design technique in the Taguchi method. World Tribology Congress 2009.. In Proc. M. low temperature 60 40 20 0 Frictional torque [mN-m] -10 8 6 4 2 0 0 20 40 60 80 100 120 Bearing temperature [ºC] Temperature 80 Frictional torque Time [hr] (c) Vacuum pressure. Nogi. It was con firmed that the retainer designed using the present technique with a high S/N ratio performed well in the bearing rotating test. For the retainer type B in Fig. Obara. the frictional torque and temperature rise several times in the environment (c). room temperature Figure. 10. REFERENCES Time [hr] (b) Atmospheric pressure. Numerical Investigation of Influence of Gravity on the Performance of Ball Bearin gs Assembled in a Space Device. similar to the retainer type A. Kyoto. 607. Maniwa. Nogi. (2001). S. In the present design technique. 9 and 10 show the results of bearing rotating tests for retainer types A and B. The horizontal axis represents time. bearing rotating tests under atmospheric pressure and a vacuum were respectively conducted to confirm their stability. Test Results Figs. Audible noises were also generated during the torque fluctuation. & Sato. For the retainer type A in Fig. Numerical Analysis of Cage Instability in Minimally Lubricated Ball Bearings. hence the present design technique is considered useful for determining the stable dynamic behavior of the ball bearing retainer. Austria. T. In Proc. Miami. IJTC2008-71154. 7. Dynamic Analysis of Minimally Lubricated Ball Bearings for Space Applications. K. Maniwa. K.2. with high and low S/ N ratios. Cotton-based phenol retainers for angular contact ball bearings with a 10 mm bore were designed using the present technique. 2.. T. Japan. whereas the vertical axis represents the frictional torque and temperature of the test bearings. respectively. 3. S. the retainer with a high S/N ratio performed well. Consequently. S. In Abstracts of papers 2nd Wo rld Tribology Congress. phenomena which are considered to indicate retainer instability. Vienna. the frictional torque appears almost completely low and stable in environments (a) and (b). whereas the clearances. (2008). & Obara. (2009). 9. 72 . To evaluate “robustness”. 596. Florida USA. Th is result indicates that t he present design technique is useful for designing high stability retainers of ball bearings. 10 Results of bearing rotating test for retainer type B with low S/N ratio 5. For two kinds of designed retainers. room temperature Bearing temperature [ºC] 80 60 40 6. However. noise factors are assigned by varying the oil film thickness slightly in numerical analysis. Time [hr] (a) Atmospheric pressure. pp 433436. STLE/ASME International Joint Tribology Conference. oil film thickness and surface roughness of the retainer are th e selected control factors.Temperature 60 40 20 0 Frictional torque [mN-m] -10 8 6 4 2 0 0 4 8 12 16 20 24 Frictional torque Bearing temperature [ºC] 80 frictional torque is low and stable in any environment.